论文部分内容阅读
针对城区激光雷达点云提出一种全自动分类方法。采用具备一定抗过分割能力的拓扑启发式分割算法对栅格高程图像进行面向对象分割;依据迭代最大类间方差(Otsu)聚类方法及两种拓扑模型实现地面图斑对象与非地面图斑对象初步分离,并合并邻接非地面对象;在地物对象中引入多次回波比率检测树木对象,采用区域面积、建筑物高度等条件区分建筑物及其他两类地物,并依据栅格索引分类。选择具有丰富地物类型的典型城区点云数据进行实验,结果表明,该算法具有良好分类精度及较强实用价值。
A kind of automatic classification method for city lidar point cloud is proposed. Object-oriented segmentation of raster elevation image is achieved by topological heuristic algorithm with some anti-segmentation ability. According to the method of Otsu clustering and two kinds of topological models, The objects were separated initially and merged with adjacent non-ground objects. The objects were introduced with multiple echo ratios to detect the trees, and the buildings and other two kinds of ground objects were classified according to the area and the height of buildings. . The typical urban point cloud data with rich feature types is selected for experiments. The results show that the algorithm has good classification accuracy and strong practical value.